pygeoprocessing.multiprocessing package

Submodules

Module contents

pygeoprocessing.multiprocessing.raster_calculator(base_raster_path_band_const_list, local_op, target_raster_path, datatype_target, nodata_target, n_workers=2, calc_raster_stats=True, use_shared_memory=False, largest_block=65536, raster_driver_creation_tuple=('GTIFF', ('TILED=YES', 'BIGTIFF=YES', 'COMPRESS=LZW', 'BLOCKXSIZE=256', 'BLOCKYSIZE=256')))[source]

Apply local a raster operation on a stack of rasters.

This function applies a user defined function across a stack of rasters’ pixel stack. The rasters in base_raster_path_band_list must be spatially aligned and have the same cell sizes.

Parameters:

  • base_raster_path_band_const_list (sequence) – a sequence containing either (str, int) tuples, numpy.ndarray s of up to two dimensions, or an (object, ‘raw’) tuple. A (str, int) tuple refers to a raster path band index pair to use as an input. The numpy.ndarray s must be broadcastable to each other AND the size of the raster inputs. Values passed by (object, 'raw') tuples pass object directly into the local_op. All rasters must have the same raster size. If only arrays are input, numpy arrays must be broadcastable to each other and the final raster size will be the final broadcast array shape. A value error is raised if only “raw” inputs are passed.

  • local_op (function) – there are elements in base_raster_path_band_const_list. The parameters in local_op will map 1-to-1 in order with the values in base_raster_path_band_const_list. raster_calculator will call local_op to generate the pixel values in target_raster along memory block aligned processing windows. Note any particular call to local_op will have the arguments from raster_path_band_const_list sliced to overlap that window. If an argument from raster_path_band_const_list is a raster/path band tuple, it will be passed to local_op as a 2D numpy array of pixel values that align with the processing window that local_op is targeting. A 2D or 1D array will be sliced to match the processing window and in the case of a 1D array tiled in whatever dimension is flat. If an argument is a scalar it is passed as as scalar. The return value must be a 2D array of the same size as any of the input parameter 2D arrays and contain the desired pixel values for the target raster.

  • target_raster_path (string) – the path of the output raster. The projection, size, and cell size will be the same as the rasters in base_raster_path_const_band_list or the final broadcast size of the constant/ndarray values in the list.

  • datatype_target (gdal datatype; int) – the desired GDAL output type of the target raster.

  • nodata_target (numerical value) – the desired nodata value of the target raster.

  • n_workers (int) – number of Processes to launch for parallel processing, defaults to multiprocessing.cpu_count().

  • calc_raster_stats (boolean) – If True, calculates and sets raster statistics (min, max, mean, and stdev) for target raster.

  • use_shared_memory (boolean) – If True, uses Python Multiprocessing shared memory to calculate raster stats for faster performance. This feature is available for Python >= 3.8 and will otherwise be ignored for earlier versions of Python.

  • largest_block (int) – Attempts to internally iterate over raster blocks with this many elements. Useful in cases where the blocksize is relatively small, memory is available, and the function call overhead dominates the iteration. Defaults to 2**20. A value of anything less than the original blocksize of the raster will result in blocksizes equal to the original size.

  • raster_driver_creation_tuple (tuple) – a tuple containing a GDAL driver name string as the first element and a GDAL creation options tuple/list as the second. Defaults to geoprocessing.DEFAULT_GTIFF_CREATION_TUPLE_OPTIONS.

Returns:

None

Raises:

ValueError – invalid input provided